Local LLM Library & Hardware Requirements

67 models you can search and filter by type, capability and size. Memory figures are working-set estimates per quantization (treat as ±) — they map model sizes to hardware tiers, not exact benchmarks. Open each model for compatible devices and a recommended build.

Best local LLMs Best coding LLMs Best RAG models LLM hardware requirements Find the right model →

Type

Capability

Size

67 models

Sort by

Browse by family

See every size in a family side by side, with the hardware each one needs.

Qwen16 DeepSeek7 Llama7 Mistral4 CodeLlama3 Gemma3 Gemma 33 LLaVA3 Phi3 StarCoder3 Granite2 Qwen2.52

On the frontier API models: Claude (Anthropic API), GPT-class (OpenAI API), Gemini-class (Google API) are listed only as quality and cost comparison anchors for the hybrid strategy — they run as hosted services, not on local hardware, and send data to the provider.

A note on honesty: model sizes, context windows and footprints change between releases, and licenses vary. Treat figures here as approximate guidance and verify the exact variant and its terms before deploying. Reasoning, vision and API entries are flagged accordingly.

Run these models inside a private AI Business OS

Run your own AI agents on hardware you control — private by design, no per-seat data leaving your premises. BrainOutput helps you pick the right machine and turn it into a working AI Business OS.

Explore the AI Business OS

DeepSeek-R1 671B (MoE)

Llama 3.1 405B

Qwen3 235B-A22B (MoE)

Qwen2.5 72B

Llama 3.1 70B

Llama 3.3 70B

DeepSeek-R1 Distill Llama 70B

Mixtral 8x7B (MoE)

CodeLlama 34B

Qwen2.5 32B

Qwen3 32B

DeepSeek-R1 Distill 32B

Qwen2.5-Coder 32B

Gemma 2 27B

Gemma 3 27B

Mistral Small 24B

DeepSeek-Coder V2 (class)

StarCoder2 15B

Qwen2.5 14B

Qwen3 14B

Phi-3 Medium (14B)

Phi-4 (14B)

DeepSeek-R1 Distill 14B

Qwen2.5-Coder 14B

CodeLlama 13B

LLaVA 13B (vision)

Gemma 3 12B

Mistral Nemo 12B

Llama 3.2 Vision 11B

Gemma 2 9B

Llama 3.1 8B

Qwen3 8B

Granite 3 8B

DeepSeek-R1 Distill 8B

LLaVA-Llama3 8B (vision)

MiniCPM-V 8B (vision)

Qwen2.5 7B Instruct

Qwen2.5 Coder 7B Instruct

Qwen2.5 7B

Mistral 7B

Qwen2.5-Coder 7B

CodeLlama 7B

StarCoder2 7B

Qwen2-VL 7B (vision)

LLaVA 7B (vision)

Gemma 3 4B

Phi-3.5 Mini (3.8B)

Llama 3.2 3B

Qwen2.5 3B

StarCoder2 3B

Gemma 2 2B

Granite 3 2B

Moondream 2 (vision)

SmolLM2 1.7B

Qwen2.5 1.5B

DeepSeek-R1 Distill 1.5B

Qwen2.5-Coder 1.5B

Llama 3.2 1B

BGE-M3 Embeddings (class)

Qwen2.5 0.5B

mxbai-embed-large (class)

Snowflake Arctic Embed (class)

Nomic Embed Text (class)

all-MiniLM (class)

Claude (Anthropic API)

GPT-class (OpenAI API)

Gemini-class (Google API)

Browse by family

Run these models inside a private AI Business OS